Using AI to measure Parkinson’s disease severity at home

Introducing an innovative artificial intelligence system designed to revolutionize the assessment of motor performance in individuals with Parkinson’s disease (PD) from a distance. In this groundbreaking approach, participants engage in a motor task involving finger tapping, all while being observed by a webcam. The resulting data, gathered from a diverse pool of 250 global participants, underwent meticulous evaluation by three distinguished expert neurologists, who adhered closely to the standards set by the Movement Disorder Society Unified Parkinson’s Disease Rating Scale (MDS-UPDRS).

The consistency and reliability of the neurologists’ evaluations were truly remarkable, showcasing an impressive intra-class correlation coefficient (ICC) of 0.88. This not only underscores the precision of their assessments but also highlights the potential of technology-assisted diagnostics in the realm of neurology. Leveraging cutting-edge computer algorithms, we harnessed this wealth of data to derive objective measurements meticulously aligned with the MDS-UPDRS guidelines. These measurements exhibited a robust correlation with the expert neurologists’ ratings, further affirming their accuracy and validity.

Through rigorous machine learning endeavors, our model, meticulously trained on these comprehensive measures, exhibited a remarkable feat – outperforming an MDS-UPDRS certified rater. The mean absolute error (MAE) achieved by our model stood at an impressive 0.59, surpassing the rater’s performance which registered an MAE of 0.79. While our model slightly trailed behind the expertise of the seasoned neurologists (with a 0.53 MAE), its proficiency remains an extraordinary leap forward in diagnostic capabilities.

A notable highlight of this methodology lies in its potential for replication across various motor tasks. By offering a framework that can be adapted to similar scenarios, we open the door to comprehensive evaluations for individuals grappling with PD and other movement disorders. This innovative approach, enabled by technology, transcends geographical barriers, making it a particularly invaluable asset in regions where access to specialized neurological care is limited.

In summation, this trailblazing AI-driven system demonstrates not only the future of motor performance assessment but also a pivotal step toward democratizing quality neurological evaluations.


Source Arxiv

Author: Neurologica